Fast Reflective Beamforming With Situational Awareness in IRS Assisted Communication Systems

In this paper, we investigate reflective beamforming methods for intelligent reflecting surface (IRS)-assisted communication systems. In our envisioned system, the IRS is equipped with computer vision to acquire situational awareness. Leveraging this information, we develop a gradient descent-based reflective beam optimization method that significantly reduces beamforming overhead. Through simulations, we demonstrate that the proposed method achieves a higher achievable rate compared to existing hierarchical codebook-based beamforming methods. Furthermore, to design prediction models for situational awareness in IRS-assisted communication systems, we investigate the impact of situational awareness estimation errors on the proposed method.